Estimating the relative contribution of parasitic infections and nutrition for anaemia among school-aged children in Kenya: a subnational geostatistical analysis

Abstract

Objectives: To quantify geographical variation in the relative contribution of parasitic infections, socioeconomic factors and malnutrition in the aetiology of anaemia among schoolchildren across Kenya, thereby providing a rational basis for the targeting of an integrated school health package.

Design: Nationally representative cross-sectional survey data were collected using standard protocols. For all included children, data were recorded on haemoglobin (Hb) concentration and common parasitic infections (Plasmodium falciparum, hookworm and schistosomes) and socioeconomic indicators. Ecological proxies of malnutrition and food security were generated using Demographic and Health Survey and UN Food and Agriculture Organization food security data, respectively. Spatially explicit, multilevel models were used to quantify impact upon child Hb concentration.

Setting: Randomly selected schools in ecologically diverse settings across Kenya.

Main outcome measures: Mean Hb concentration adjusted for infection, nutritional and socioeconomic risk factors; associated risk ratios and adjusted Population Attributable Fractions (PAFs) for anaemia, by region.

Results: Data were available for 16 941 children in 167 schools; mean Hb was 122.1 g/l and 35.3% of children were anaemic. In multivariate analysis, mean Hb was significantly lower in boys and younger children. Severe malnutrition and interactions between P falciparum and hookworm infections were significantly associated with lower Hb, with greater impacts seen for coinfected children. The contribution of risk factors to anaemia risk varied by province: in 14-year-old girls, PAFs ranged between 0% and 27.6% for P falciparum, 0% and 29% for hookworm and 0% and 18.4% for severe malnutrition.

Conclusions: The observed geographical heterogeneity in the burden of anaemia attributable to different aetiological factors has important implications for the rational targeting of antianaemia interventions that can be included in an integrated school health programme.

Figure 1

Schematic of adopted analysis strategy. The dual aim of this analysis strategy was to (1) model the impact of risk factors on the children's haemoglobin (Hb) levels, and (2) estimate the adjusted Population Attributable Risk Fraction (PAF) associated with important, modifiable risk factors. This process was repeated at each realisation of the Bayesian hierarchical model to provide posterior mean estimates and credible intervals for all parameters of relevance.

Figure 2

Spatial distribution of mean haemoglobin distribution across Kenya, by school. Hb is adjusted for altitude; population density (persons/km²) is shown for reference.

Figure 3

Graph of spatial random-effect (school-level residual) from the multivariate, Bayesian hierarchical model for Hb. Each point represents one school; the posterior mean is shown by the dot and the 95% BCI are represented by the error bars. The horizontal straight line in the middle of the graph represents the overall mean of the residuals, which are zero centred. Inset shows the school locations, shaded by the value of the school-level residual.